K-Means algorithm-based detection for wavelength division multiplexed OOK PD-NOMA system over turbulent optical channel

Abstract

We have experimentally demonstrated the efficacy of k-means algorithm-based detection scheme in a two-user-based wavelength division multiplexed (WDM) on-off keying (OOK) power-division non-orthogonal multiple access (PD-NOMA) system over turbulent free-space optical (FSO) channels. The proposed system utilizes both power and wavelength division multiplexing to enhance the capacity of the transmission system. The proposed system also incorporates successive interference cancellation technique that utilizes unsupervised machine learning technique, which is K-means clustering, to detect the signal. Two different weak turbulence conditions in the free-space optical channels have been realized in the laboratory environment having link lengths of 2.2 m (near user) and 3.2 m (far user). The results show that the measured probability density function of intensity fluctuation closely matches the log-normal distribution. The experimental results also show that the K-means-based detection provides the BER, which is well below the Pre-FEC limit. The maximum power penalty for the near (far) user is found to be 3.1 dB (4.2 dB) at a BER of 10 − 3 for weak turbulence conditions. The proposed and implemented k-means-based detection scheme for the WDM-OOK PD-NOMA system can also be used for FSO communication systems.